Skip to main content

Full text of "Weak convergence of orthogonal polynomials"

See other formats


G^ ■ Walter Van Assche 

Katholieke Universiteit Leuven 

Abstract. The weak convergence of orthogonal polynomials is given under conditions on the asymptotic 
behaviour of the coefficients in the three-term recurrence relation. The results generalize known results and 
are applied to several systems of orthogonal polynomials, including orthogonal polynomials on a finite set of 

1. Introduction 

Let Pn{x) be a system of orthonormal polynomials on the real line, with orthogonality measure /x, i.e. 
/x is a probability measure for which all the moments exist and 



g. I Pnix)pmix) dfiix) = 5m,n: m, n > 0. 

^ . When the support of fx consists of a finite number of points Xi, 0:2, . . . , Xn, then we only consider th 

Tjj- ! polynomials up to degree N and Pn{x) has its zeros at the support {a;i,a;2, . . . ^x^}. It is well knowi 

^ \ that orthonormal polynomials satisfy a three-term recurrence relation 

t^ : (1.1) XPrr{x) = ari+lPn+l{x) + hnPn{x) + a„p„_i(x), 71 > 0, 


•• . with initial values Pq{x) = 1 and p-i{x) = 0. Here 
\^ ■ an+1 = / xpn{x)pn+i{x) diJi{x), h^ = j xp^(x) (i/i(x) G M, n > 0. 

Usually the orthonormal polynomials are chosen in such a way that the leading coefficient 7^ = (aia2 • ■ ■ a, 
is positive, and then a^ = 7n/7n-i > for every n > 1. 

We are interested in the weak asymptotic behaviour of the orthonormal polynomials ^^(2^)- This mean 
that we want to investigate the behaviour for n -^ 00 oi integrals of the form 

f{x)pl{x)d^i{x), feCb 

where C^ is the linear space of bounded and continuous functions on M. 

We will consider a one-parameter orthogonality measure Hk {k G N), the parameter being discrete 
This implies that the recurrence coefficients and the orthogonal polynomials all depend on this discret( 
parameter and we write 

(1.2) xpn{x] fXk) = an+i,kPn+i{x; fXk) + bn,kPn{x; fXk) + an,kPn-i{x; Hk), n > 0. 

Our main result will be the limit as n -^ 00 of integrals of the form 


The special case k = I = then gives the desired weak convergence, but the general case with k,l E 2 
also has useful applications: these integrals are related to the transition probabilities of birth and deatl 
processes and random walks. 

Such one-parameter families do occur frequently in various applications and limiting procedures. Fo 
example the rescaling of orthogonal polynomials 

Pn{ckx;^) =pn{x;^k) 
gives the one-parameter family of measures Hk with distribution functions satisfying 

^k{t) = ^i{ckt), t e R. 

Other examples include orthogonal polynomials in which some of the parameters are allowed to ten( 
to infinity together with the degree, e.g., if Pn' (x) is the Jacobi polynomial of degree n, with weigh 

w{x) = {l-x)'^{l + xf, -1<X<1, 

then Pn " '^' (x) is an orthogonal polynomial of degree n with weight function 

Wk{x) = w''{x){i - xy{i + xY 

-1 < X < 1. 

The main result will be in terms of a doubly infinite Jacobi matrix 

/•• •• \ 


QjQ On Qj-^ 

a? 6? 4 

aO 60 a« 


••• ••• •••/ 

and is the following 

Theorem. Suppose that the recurrence coefficients in (1.2) satisfy 

(1.3) lim an+k,n = a° > 0, lim bn+k,n = bl e 

n — >oo n — >cxD 

for every k E Z,. Then 

lim / f{x)pn+k{x] Hn)Pn+l{x; Hn) dHn{x) 

f{x){Ak{x) Bk{x))dii{x) 


for every polynomial f . Here n is the spectral matrix of measures for the doubly infinite Jacobi matrix C 
with entries a^, 6° (n G Z) and 




for n >0, 
for 71 < 0, 

„..A-H, ^ t^\ „^J ^ t^\ +U^ ^^-l-U, 

,„7 ^^/..^^™.-^/„ n..n + U 

-.■PQ ^A ^^4-r 

■,-^A..^U. „0 UO I. 


2. Spectral theory for Jacobi operators 

If we put the coefficients a^ > (n = 1, 2, 
(1.1) in a tridiagonal matrix 

and bji G M (n = 0, 1, 2, . . . ) of the recurrence relatioi 



/bo ai 
ai hi 







then J is the Jacobi matrix associated with these orthogonal polynomials. If // is supported on A^ points 
then a„ = = 6^ for n > A^ and J is a, N x N matrix with eigenvalues at the support {xi, X2, . . . , xn} 
Note that 


and by induction we find 



xpi{x)pj{x) dn{x) 



X pi{x)pj{x) dii{x] 


In fact, J : £2(^5 C) -^ £2(^5 C) acts as a linear operator on the Hilbert space £2(^5 C) = {V' : V'i ^ 
C and Xli^o IV'iP < 00}. The operator is symmetric on the initial domain consisting of finite linea 
combinations of the basic vectors {e+ = (0, 0, . . . , 0, 1, 0, . . .) : n > 0} and by imposing appropriati 

n zeros 

conditions on the recurrence coefficients an+i,6n (e.g., boundedness) this operator can be extended ii 
a unique way to a self-adjoint operator on the maximal domain {-!/; G £2(1^, C) : J'^ G £2(1^, C)}. Thi 
operator has a cyclic vector, i.e., if we take e^ = (1,0,0,0,...), then the linear span of {J^Cq : k = 
0, 1, ... } is dense in £2(N, C). The spectral theorem (see, e.g., Akhiezer and Glazman [2] or Stone [19] 
then implies the existence of a measure n and a linear mapping A : £2(N, C) -^ -^2(a*) with Ae^ = 1 anc 
AJt/j = MAt(j for every t(j G £2(N, C), where M is the multiplication operator 


The mapping A is unitary, meaning (Atp^Acj)) ~ [ip^cj)). The mapping A thus maps e^ to the constan 
function t 1— *> 1, Je^ to the identity t ^^ t^ and in general A maps J^e^ to the monomial t ^-^ f^. Heno 
the fact that e^ is a cyclic vector is equivalent with the density of polynomials in L2 (n) ■ By inductioi 
and by using the recurrence relation (1.1) we see that A maps the basic vector e^ to the polynomial Pn 
The unitarity thus implies that 

Pn{t)Pm{t) d^{t) = (e+,e+) = S^ 

which shows that the spectral measure n for the operator J is the orthogonality measure for the orthogona 
polynomials Pn{x) {n = 0,1,2,...). These elements from spectral theory, applied to the semi-infiniti 
Jacobi matrix J, are well-known (see e.g., Akhiezer [1], Dombrowski [6], Sarason [18], Stone [19]). 
We will also need to use doubly infinite Jacobi matrices of the form 

/■■. ■■. 



with Ofc > 0, 6fc G M for k E Z. Again these are operators, now acting on the Hilbert space £2(^7 C) = {i/j 
ifji E C and Yl'iZ-00 l'*/'*!^ < ^^}- "^^^ spectral theory of such matrices is less known, but has also beei 
developped (see, e.g., Berezanskii [4], Nikishin [17], Masson and Repka [13]). We will briefly recall souk 
of the elements, which we have taken from Nikishin [17] (see also Berezanskii [4, Chapter VII, §3]). W 
will flrst consider bounded matrices JT", i.e., we assume that a^ and bn (n G Z) are bounded sequences 
It turns out that it is convenient to introduce the 2x2 matrices 


b-1 ao 
«o ^0 







n = l,2, 

We can now study the semi-inflnite Jacobi block matrix 


/Bo Ai \ 

Ai Bi A2 

A2 B2 As 

A3 S3 A4 

which contains 2x2 matrices. As an operator it acts on £2(N, C^) in the sense that J't/j for t/j G £2(Z, 
corresponds to J\l/ with \1/ G £2(N, C^) given by 


*n = ( ":r' 1, n = o,i,2. 

In this way we have transformed the study of the doubly-inflnite Jacobi matrix to the study of a semi 
inflnite Jacobi block matrix. Such semi-infinite block matrices are closely connected to orthogonal matri: 
polynomials, in the same way as ordinary Jacobi matrices are connected to scalar orthogonal polynomial 
(see, e.g., Aptekarev and Nikishin [3]). 

Consider the standard basis e„ (n G Z) in £2(Z, C), i.e., 

(Cn)i = Oi^rn t,n E £, 

then the linear span of {J'^e-i, J^e^, /c, / G N} is dense in £2(Z, C). By the spectral theorem there exist 
a matrix-measure 

Afl,2 IJ'2,2 


and a unitary linear mapping A : £2(Z, C) -^ -^2(1^) with 

Ae_i = I ^ 1 , Aeo 

such that 

AJij = MAV', 

where now M is the multiplication operator in the space L2{p) of vector valued functions. The inne 
product in the space L2{n) is given by 


whereas J'^e^ is mapped to 

Let J"*" be the semi-infinite Jacobi matrix 












/b-i a-i 
a-i b-2 




b-3 a_3 



and similarly J be the semi-infinite Jacobi matrix 


We denote by Pn{x) the orthonormal polynomials corresponding to the Jacobi- matrix J^ satisfying th 
recurrence relation 

(2.6) XPn{x) = an+lPn+l{x) + bnPn{x) + anPn-lix), 

with initial values p-i(x) = 0,po{x) = 1, and by qn{x) the orthonormal polynomials for J~ satisfying 

(2.7) XQnix) = a-ri-iqn+iix) + b-ri-iQuix) + a_„Q„_i(x), 

with initial values Q-i(x) = 0,^0(3^) = 1- We will show, by induction, that for n G N the mapping J 
maps the basis vector e„ to Ae^ which is the vector function 


and the basisvector e-n to Ae_n given by 



t h^ 

._ao_^(l) (f) 

Here pn (x) and Qn (x) are the associated polynomials, i.e., the orthonormal polynomials corresponding 
to the Jacobi matrices J+ and J~, with the first row and column deleted. This is clear for e_i and cq 
Assume that this is true for < n < k, then from 

Jek = ak+iCk+i + bkCk + akCk-i 

it follows that 

AJ^Cfc = ak+iA.ek+1 + bkA-Ck + afcAcfc-i, 
and since KJck = MKck this gives 


and KJe-k — MAe-k, this gives 

From the unitarity we find for m,n E 7^ 

Hence the matrix polynomials 


J^^n\t) — I ^ i^i^ 


J Pn{t)dl^{t)PUtr = Q J") 5^,^. 

Therefore these matrix polynomials are orthonormal with respect to the matrix-measure //. Note tha 
these matrix polynomials satisfy 

1 0' 

SO that they are the orthonormal polynomials corresponding to the block Jacobi matrix J given in (2.5) 
The polynomials Pn{x) (n = 0, 1, 2, . . . ) are orthonormal with respect to some probability measure /i"*" 
and the polynomials qn{x) {n = 0, 1, 2, . . .) are orthonormal with respect to some probability measur( 
|U~. In order to find a relation between the matrix-measure ^ and the measures /U"'" and /U~, we observ( 
that the unitarity of A implies 

{{z - J)~^en,em) = / -{Aen)*d^{t)Aem, 

J z-t 

where {z — J)~^ is the resolvent of J, which is well-defined for every z outside the spectrum of J . Sinci 
J is symmetric and bounded, it follows that J is self-adjoint so that its spectrum is a subset of the rea 
line. The Stieltjes transform of the matrix-measure ^ is determined by 

{{z- J)~^e-i,e-i) = I -d^i^i{t), {{z- J)~'^e-i,eo) = / -<i;Ui,2(t), 

J z-t J z-t 

{{z- J)~^eo,eo) = j ——-dii2,2{t). 
On the other hand, if we write 

{z- J)~^eo = r ={... ,r_2,r_i,ro,ri,r2,...), 
then [z — J')r = eo, which gives the infinite system of equations 


combination of the orthogonal polynomials Pn{z) (respectively qn{z)) and the functions of the secom 
kind Pn{z) (respectively qn{z)) given by 

Pn{z) = / — (i/i+(a:), qn{z) = / — d^~{x), 

J z — X J z — X 

with aoP-i{z) = 1 = aoq-i{z). These functions of the second kind have the property that they are i 
minimal solution of the recurrence relation, and for z G C\R they satisfy lim^^oo Pn{z) = lim^^oo qn{z) = 
0. The fact that r G £2(^7 C) thus implies that r^ and r^^ are (up to a constant factor) given by th 
functions of the second kind Pn{z) and qn-i{z) respectively. The constant multiple is determined h 
setting n = 0, giving 

Tn = To ^ , r-ri = aoroQn-i(^), U > 0. 

In particular ri = ropi{z)/po{z) and r_i = aoroqo{z). Inserting in (2.10) gives 



z - alqo{z) -bo- aipi{z)/po{z) ' 

which by using aipi{z) = (z — bo)po{z) — 1 gives 

^ Pojz) 


In a similar way we may investigate 

(Z - jy^e-x = S ==(... , S_2, S_i, So, Sl, S2, ■■■), 

which gives the infinite system of linear equations 

^Sjt = ttfcSfc-i +6A:Sfc +aA;+lSA:+l, A: > 0, 

(2.11) ^s-jt = a_fcS_fc_i + 6_fcS_fc + a_A;+is_fc+i, /c > 2, 

zs-\ - 1 = a_is_2 + h-\S-\ + aoSQ. 

Now we find 

~ / N qu-\{z) 

Sn=aoS-xPn{z), S-n = S-1 ^ . . , 

90 (^) 

and inserting this in (2.11) gives 


z - a-iqi{z)/qo{z) - b-i - alpo{z) ' 
which by using a-iqi{z) = [z — b-i)qo{z) — 1 becomes 

^ qojz) 

The Stieltjes transform of the matrix of measures fu. is thus given by 

1 , /,^ qo{z) /" 1 ^ m Po(^) 

z-t ' 1 - a^po{z)qo{z) J z-t ' 1 - a5po(^)<7o(2;) 

walter van assche 

3. Proof of the theorem 
The first important observation is that for f{x) = x"^ one has 

x'^Pn+kix] Hn)Pn+lix] Hn) dHn{x) = ( 7^6+^^, e+_^^) , 

where Jn is the semi-infinite Jacobi matrix with the recurrence coefficients ak,n,bk,n, {k = 0,1,2,...) 
Consider the semi-infinite operator J^ as a doubly infinite Jacobi matrix by taking {Jn)i,j = wheneve: 
i < or J < 0. If 5" is the shift operator on £2(^7 C) acting as Sek = ejt+i, then 

One easily verifies that this expression is given by 

(('-' ) ^n^ ^ki^l) = / ^ {'Jn)n+k,n+k+ii{'Jn)n+k+ii,n+k+ii+i2 

11,12, ■■■ ,im~lE{-l,0,l} 

il+i2-\ \-im-i+k-l e{-l,0,l} 


n+k+ii+i2-\ \-im-i,n+l- 

This is a finite sum, containing the matrix entries {Jn)n+r,n+s where r and s remain bounded. Th 
hypothesis (1.3) implies that {S*)'^JnS'^ converges entrywise to JT", where J' is the doubly infinite Jacob 
matrix containing the coefficients a^, 6^, {n E Zi). The hypothesis (1.3) thus implies that 

lim / x'^pn+kix; iJn)Pn+i{x; Hn) diJn{x) 

n — >oo J 

/ J >Jk,k+ii>Jk+ii,k+ii+i2 ' ' ' '-^k+ii+i2-\ Hm-l,l^ 

11,12, ■■■ ,im-l6{ — 1,0,1} 
il+i2 + ---+im-l+k-l e{-l,0,l} 

and the latter expression is equal to 


Therefore we find that (S*)'^ J^ S"^ converges entrywise to J"^. By the unitarity of the mapping A 
£2(2, C) -^ -^2 (a*) 5 transforming the action of J' on £2(^5 C) to the action of the multiplication operato 
M on L2(/u), we have 

iJ'^ek^ei) = J t'^{Aekrdijit)Aei, 

and thus the theorem for f{x) = x"^ (and hence for every polynomial /) follows from (2.8) and (2.9). 

In order to proof the theorem for every / G C^, we consider the linear operator Hn — {S*)'^ JnS"^ . W 
would like to prove that f{Hn) = {S*)'^f{Jn)S^ converges weakly to f{J)^ since then 

lim {{STf{Jn)S'^ek,ei) = {f{J)ek,ei), 

n— >oo 

and this is precisely the weak convergence stated in our theorem. Observe that HnCk = ak+n,nek-i H 
bk+n,nek + ak+n+i,nek+i whenever k + n > 0, hence the condition (1.3) implies the convergence of HnCi 
to J'ek = a^Ck-i + h^Ck + a^ i ^e^+i for every k E Z. All finite linear combinations of the basis element; 


4. Examples 

The class M{a, b). The class M{a, b) consists of all orthogonal polynomials Pn{x] /u) (or all probabilit;' 
measures //) with recurrence coefficients that satisfy 

lim ttn = a/2, lim bn = b. 

n — >oo n — >oo 

We can apply the theorem with the family of measures ^k = A*? i-^-, with all the orthogonality measure; 
the same. If a > then we can, without loss of generality, only consider M(1,0). The doubly infiniti 
Jacobi matrix J then consists of on the diagonal and 1/2 on the subdiagonals. The semi-infinite Jacob 
matrices J^ and J~ are the same and the corresponding orthogonal polynomials are the Chebyshe'' 
polynomials of the second kind, which are orthogonal with respect to the measure (2/7r)Vl — x"^ dx oi 
the interval [—1, 1]. The Stieltjes transform of this measure is 

f>o{z) = qo{z)=2[z-^z^-l]. 
Therefore the Stieltjes transform of the spectral matrix of measures for the Jacobi matrix J is given b^ 

diii^i{x) = = / dn2,2{x), 

z — x ' ^ z^ — 1 J z — x 

1 , , , z-V^^l 


Z - X ' V-2^ - 1 

from which one easily finds that the spectrum of JT is [—1, 1] and 

dni^i{x) = dn2,2{x) = - , dni^2{x) 


TT \/l^^X^ ' TT Vl — X 

From our theorem we thus find 

//■ 1 /"^ f(x) 

f{x)pl{x;^)d^{x) = / f{x)dn2,2{x) = - dx, 

J TT J_i VI - x^ 

and in general 

lim f{x)pn{x;^)pn+k{x;^)d^{x) = - f{x) , dx 

n^oo J T^ J-1 vl-a;^ 

= - f f{x)^^^^dx. 

Here we have used the identity 

Uk{x) - xUk-i{x) = Tfc(x), 

where Tk{x) is the Chebyshev polynomial of the first kind. This result is well-known and can already bi 
found in [15, Theorem 13 on p. 45]. See also [20, Theorem 2 on p. 438]. 

Unbounded recurrence coefficients. Suppose that we have a sequence of orthogonal polynomial 
Pn{x; fj.) satisfying the three-term recurrence relation (1.1). If we rescale the variable by a positiv 
and increasing sequence Ck and consider the one-parameter family of polynomials Pn{ckX; n), then thesi 
polynomials satisfy a recurrence relation of the form (1.2) with 



lim ^^ = 1, 

then it is easy to show that 

hm an+k,n = a/2, hm bn+k,n = b. 

n — >cx3 n — >oo 

Therefore the conditions of our theorem are vahd and the theorem is true, with J' the doubly infiniti 
Jacobi matrix with constant entries b on the diagonal and a/2 on the subdiagonal. The orthogona 
polynomials for the Jacobi matrices J~^ and J~ are Un{^^) and the matrix of measures is supported oi 

[b — a,b + a] and given by 

1 dx 1 (x — b) dx 

TT ^a2 -{x- 6)2 aTT ^a^ - {x - by 

We thus have 

1 f^^"" fix) 

lim / f{x/cr,)pl{x]n)dij{x) = I f{x) diX2,2{x) = - ===dx. 


b-a ^Jo? - {X - 6)2 

This result can already be found in [10, Lemma 1 on p. 52] and [16, Lemma 3 on p. 1188]. In general W( 

hm f{x/Cr,)Pn{x;iJ)Pn+k{x]H)dn{x) = - f{x)^====^==dx. 

"^°° J TT Jb_„ ^a^ - (x - 6)2 

Wall polynomials. The orthonormal Wall polynomials Wn{x]b,q), with (0 < g < 1, < 6 < 1) ar 
orthogonal on the geometric sequence {q"', n = 1, 2, 3, . . . } and have recurrence coefficients 


,(6,Q)=Q"V6(l-Q-)(l-6g-i), bn{b,q) = q^[b + q-{l + q)bq^]. 

If we consider the Wall polynomials Wn{x] 6; c^' ^), where < c < 1, then we have a one-parameter famib 
of orthogonal polynomials with recurrence coefficients 

an,k = an{b, c^/^), br,,k = 6n(6, c^/''), 

and one easily verifies that 

lim an+k,n = ca/6(1 -c)(1 -be) = A/2, lim bn+k,n = (6 + 1 - 26c)c = B. 

n — >o<D n — -TOO 

Hence we can apply the theorem, where JT again is a doubly infinite Jacobi matrix with constant entrie; 
B on the diagonal and A/2 on the subdiagonals. Therefore again we have the asymptotic behaviour ii 
terms of Chebyshev polynomials (first and second kind) as in the previous two examples. These threi 
examples are all covered by Theorem 2 in [21, p. 307] which covers the special case of our theorem witl 
a doubly infinite Jacobi matrix with constant entries. In particular this asymptotic behaviour for Wal 
polynomials was used in [21] to show that the product formulas for Legendre polynomials are a limitin) 
case of the product formulas for little g-Legendre polynomials as g — * 1. 

Jacobi polynomials Pn "' ^ (x). The Jacobi polynomials Pn \x) are orthogonal on [—1, 1] witl 
the weight function {1 — x)°' {1 + x)^ . The orthonormal Jacobi polynomials 

„(a,/3). N _ / 2n + a + /J + l n!(a + /3 + 2)^ pi-'^P^Jx) 
^" ^^^ \ n + a + 3 + 1 ia+DJB + Dr,'^^ ^^^ 


If we consider the Jacobi polynomials pn "' {x) with a > 0, 6 > 0, then the recurrence coefficienti 

are a^^k = a^^ak + a, 6/c + 13) and h^^k = bn{ak + a,bk + (3), and one easily finds 

2J{a + l){b + l){a + b + l) ^. ^ b^-a^ 

hm a^+fc.n = — ^^ 7 — , , , ^,o , hm bn+k, - 

(a + 6 + 2)2 ' n^oo "+'^'" (a + 6 + 2)2- 

Hence our theorem applies again with a constant doubly infinite Jacobi matrix J'. This result has no 
been given in the literature, but complements the known results concerning strong asymptotics and zer( 
behaviour given in [5] [9] [14]. 

Laguerre polynomials L^'^+"(nx). The Laguerre polynomials L" are orthogonal on [0, oo) with weigh 
function x'^e~^ . The orthonormal Laguerre polynomials 

have recurrence coefficients 

«n(tt) = \/n{n + a), bn{a) = 2n + a + 1. 

If we consider the polynomials p'^^~^°'{kx), then we have an,k = an{ak + a)/k and bn,k = bn{ak + a)/k 
hence one easily finds 

lim an+k,n = a + 1, lim bn+k,n = a + 2, 

n — »oo 

SO that once more our theorem can be applied with a constant Jacobi matrix J'. This complements th( 
asymptotic behaviour for such Laguerre polynomials given in [5] and [8]. 

Dual Hahn polynomials. The dual Hahn polynomials Rn{x) — Rn{x; a, (3, N) are given by the recur 
rence relation 

-xRk{x) = DkRk-i{x) - [Dk + Bk)Rk{x) + BkRk+i{x), 

with initial condition Ro{x) = 1 and R-i{x) = [11]. Here 

Bk = {N -l-k){a + l + k), Dk = k{N + P-k), 

and the polynomials are orthogonal on the quadratic lattice {xk = k{k + a+(3 + l) : /c = 0, 1, 2, . . . , A^ — 1 
with weights 

TTk = 7rk{a,(3,N) 

'N - 1\ T{P + N) r{k + a + l)r{k + a + (3 + l] 

k )r{N + a + P + k + l) r(/c + /3 + l)r(a + l) 

{2k + a + (3 + l] 

at these points Xk, so that these polynomials are only defined up to degree A^. The orthonormal polyno 
mials are 

Pn{x;a,/3,N) = ^~^ -— Rn{x;a, (3,N), 

(ce+n\ 1/^ /f3+N-l-n\ ^1 ^ 
\ n I \ N-l-n ) 

with recurrence coefficients a2 = DnBn-i and bn = D^ + B^. These polynomials are useful in th 
description of a genetic model of Moran, as is worked out in [111 [121 and [71. Consider the nolvnomial 


One easily finds 

2 r for /c > 0, r for /c > 0, 

^^oo '^+'=''^ \ _A;(/3 - k) for A; < 0, n^oo "+'='" \ _2fc + /3 - 1 for A; < 0. 

The doubly infinite Jacobi matrix J' corresponding to these asymptotic formulas is therefore only a semi 
infinite Jacobi matrix which coincides with the Jacobi matrix J~, for which the corresponding orthogona 
polynomials are the Laguerre polynomials L(^{x). The spectral matrix of measures of J^ thus reduces t( 
iU2,2 = = /Ui,2 and dni^i{x) = x^e~^ /T{(3 + 1) on [0, oo). Our theorem can easily be proved also for th 
case where JT reduces to a semi-infinite Jacobi matrix, and we thus have 

lim V/(xj/n)p2_^(x^.;Q,,/^,n)7rj-,^= / /(x) dx, 

where Xj = j{j + a + (3 + 1) and TVj^n = T^jict, /5, n). In general we have for /c, / > 1 

n — >oo 


lim y^ f{xj/n)pn-kixj; a, /?, n)pn-i{xj;a, (3, n)Tij 

\^) / i-/ Nr/3 / \tI3 I \ ^ ^ 


i-i Jo 

J f{x)Ll_^{x)L^_^{x) ^ ^ dx, 

where hk = { ~l ) is the norm of L^{x). 

If we consider the polynomials Pn{k^''^x + /c^/2; a, /c/2, /c), then the recurrence coeflacients are zero fo: 
n > k and for n < k we have 


2 n{k — n){3k/2 — n){a + n) 

,k= p ' 

{k-n-l){a + n + l)+ n{3k/2 - n) - k^/2 

Now one easily finds 

for /c > 0, 

-A;/2 for /c < 0, 

lim a^_^fc = <j " ^ ^^ r ," . r 1™ ^n+fc,n = for /c G Z. 

n — ■roo 

The doubly infinite Jacobi matrix J' again coincides with J , which is now the semi-infinite Jacobi matri: 
for the Hermite polynomials Hn{x). We thus find 

lim y]f( ^ 3/2 — Pn-i(a^i;«.'^/2,?^)7rj,n= / f{x)^^dx, 

"^°°^ V ^/ / J-oo VTT 

where now tTj ,^ = TVj{a, n/2, n), and in general 



/x- - n^/2\ 

> f I — :r-n I »„ _ I, f.T „■ I O. ?7, /2. ?7,1»„ _; f.T™- 1 «. ?7, /2. ?7, W„- 

orthogonal polynomials 1: 


1. N. L Akhiezer, The Classical Moment Problem, Oliver and Boyd, Edinburgh, 1965. 

2. N. L Akhiezer and L M. Glazman, Theory of linear operators in Hilbert space, vol. I, Pitman, Boston, 1981. 

3. A. I. Aptekarev and E.M. Nikishin, The scattering problem for a discrete Sturm- Liouville operator. Mat. Sb. 121 (163 
(1983), 327-358; Math. USSR Sb. 49 (1984), 325-355. 

4. Yu. M. Berezanskii, Expansions in Eigenf unctions of Self- Adjoint Operators, Amer. Math. Soc, Providence, RI, 1968 

5. L-C. Chen and M. E. H. Ismail, On asymptotics of Jacobi polynomials, SIAM J. Math. Anal. 22 (1991), 1442-1449. 

6. J. Dombrowski, Orthogonal polynomials and functional analysis. Orthogonal Polynomials: theory and practice (F 
Nevai, ed.), NATO-ASI series C, vol. 294, Kluwer, Dordrecht, 1990, pp. 147-161. 

7. B. Dunham, Fluctuation theory for Moran's model in genetics, manuscript. 

8. W. Gawronski, Strong asymptotics and the asymptotic zero distributions of Laguerre polynomials L„ " " and Hermit 
polynomials ffi""+"\ Analysis 13 (1993), 29-67. 

9. W. Gawronski and B. Shawyer, Strong asymptotics and the limit distribution of the zeros of Jacobi polynomial 
Pn^ a, n p)^ Progress in Approximation Theory (P. Nevai and A. Pinkus, eds.). Academic Press, Boston, 1991 
pp. 379-404. 

10. J. S. Geronimo and W. Van Assche, Relative asymptotics for orthogonal polynomials with unbounded recurrence coe] 
ficients, J. Approx. Theory 62 (1990), 47-69. 

11. S. Karlin and J. L. McGregor, The Hahn polynomials, formulas and an application, Scripta Math. 26 (1961), 33-46. 

12. , On a genetic model of Moran, Proc. Cambridge Philos. Soc. 58 (1962), 299-311. 

13. D. R. Masson and J. Repka, Spectral theory of Jacobi matrices in £"^(2,) and the su{l, 1) Lie algebra, SIAM J. Matb 
Anal. 22 (1991), 1131-1146. 

14. D. Moak, E. Saff and R. Varga, On the zeros of Jacobi polynomials P""''^"(a;), Trans. Amer. Math. Soc. 249 (1979) 

15. P. G. Nevai, Orthogonal Polynomials, Memoirs Amer. Math. Soc, vol. 213, Amer. Math. Soc, Providence, RI, 1979. 

16. P. G. Nevai and J. S. Dehesa, On asymptotic average properties of zeros of orthogonal polynomials, SIAM J. Math 
Anal. 10 (1979), 1184-1192. 

17. E. M. Nikishin, Discrete Sturm- Liouville operators and some problems of function theory, Trudy Sem. Petrovsk. 1( 
(1984), 3-77; J. Soviet Math. 35 (1986), 2679-2744. 

18. D. Sarason, Moment problems and operators in Hilbert space. Moments in Mathematics, Proceedings of Symposia ii 
Applied Mathematics, vol. 37, Amer. Math. Soc, Providence, RI, 1987, pp. 54-70. 

19. M. H. Stone, Linear Transformations in Hilbert Space, Amer. Math. Soc. Colloq. Publ., vol. 15, Amer. Math. Soc. 
Providence, RI, 1932. 

20. W. Van Assche, Asymptotics for orthogonal polynomials and three-term recurrences. Orthogonal Polynomials: theor 
and practice (P. Nevai, ed.), NATO-ASI series C, vol. 294, Kluwer, Dordrecht, 1990, pp. 435-462. 

21. W. Van Assche and T. H. Koornwinder, Asymptotic behaviour for Wall polynomials and the addition formula for littl 
q-Legendre polynomials, SIAM J. Math. Anal. 22 (1991), 302-311. 

22. J. Weidmann, Linear Operators in Hilbert Space, Graduate Texts in Mathematics, vol. 68, Springer- Verlag, New York 

Department of Mathematics, Katholieke Universiteit Leuven, Celestijnenlaan 200 B, B-3001 Heverlei 

E-mail address: